{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,11]],"date-time":"2026-09-11T05:06:59Z","timestamp":1789103219864,"version":"build-2803163510"},"reference-count":31,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,9,11]],"date-time":"2026-09-11T00:00:00Z","timestamp":1789084800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Artif. Intell."],"abstract":"<jats:p>Financial entity-level sentiment analysis aims to identify sentiment polarity toward specific financial entities in financial texts. This task is challenging because a single sentence may contain multiple entities with opposite sentiment orientations, while financial sentiment is often conveyed through domain-specific and implicit cues, increasing the risk of entity confusion. To address these challenges, we propose MSAR, a Multi-Stage Adversarial Representation Learning framework built on a FinBERT-CRF backbone. At the textual level, a Model-Sensitive Metric generates adversarial examples by jointly considering model sensitivity, semantic consistency, and linguistic fluency. At the representation level, Adaptive Adversarial Regularization dynamically adjusts the regularization strength according to gradient sensitivity and output-distribution variation. By integrating discrete textual perturbation with continuous representation regularization, MSAR enhances robust entity-level sentiment prediction. Experiments on FinEntity and SentFin show that MSAR outperforms representative baselines. Ablation, sensitivity, and low-resource experiments further confirm the effectiveness and robustness of the proposed components. These findings demonstrate that jointly optimizing textual and representation-level robustness improves financial entity-level sentiment analysis, particularly under low-resource conditions.<\/jats:p>","DOI":"10.3389\/frai.2026.1888120","type":"journal-article","created":{"date-parts":[[2026,9,11]],"date-time":"2026-09-11T04:31:18Z","timestamp":1789101078000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["MSAR: multi-stage adversarial representation learning for robust financial entity-level sentiment analysis"],"prefix":"10.3389","volume":"9","author":[{"given":"Xinwen","family":"Wan","sequence":"first","affiliation":[{"name":"School of Internet, Anhui University","place":["Hefei, Anhui, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Junjie","family":"Zhao","sequence":"additional","affiliation":[{"name":"School of Internet, Anhui University","place":["Hefei, Anhui, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zheng","family":"Liang","sequence":"additional","affiliation":[{"name":"School of Internet, Anhui University","place":["Hefei, Anhui, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Aoxue","family":"Shen","sequence":"additional","affiliation":[{"name":"School of Internet, Anhui University","place":["Hefei, Anhui, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Wenwen","family":"Chen","sequence":"additional","affiliation":[{"name":"School of Internet, Anhui University","place":["Hefei, Anhui, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Andi","family":"Xiao","sequence":"additional","affiliation":[{"name":"School of Internet, Anhui University","place":["Hefei, Anhui, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,9,11]]},"reference":[{"key":"B1","article-title":"Better fine-tuning by reducing representational collapse","author":"Aghajanyan","year":"2020","journal-title":"arXiv preprint arXiv:2008.03156"},{"key":"B2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D18-1316","article-title":"\u201cGenerating natural language adversarial examples,\u201d","author":"Alzantot","year":"2018","journal-title":"Proceedings of the 2018 conference on empirical methods in natural language processing"},{"key":"B3","article-title":"FinBERT: financial sentiment analysis with pre-trained language models","author":"Araci","year":"2019","journal-title":"arXiv preprint arXiv:1908.10063"},{"key":"B4","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2020.acl-main.777","article-title":"\u201cSeqVAT: virtual adversarial training for semi-supervised sequence labeling,\u201d","author":"Chen","year":"2020","journal-title":"Proceedings of the 58th annual meeting of the association for computational linguistics"},{"key":"B5","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/P18-2006","article-title":"\u201cHotFlip: white-box adversarial examples for text classification,\u201d","author":"Ebrahimi","year":"2018","journal-title":"Proceedings of the 56th annual meeting of the association for computational linguistics"},{"key":"B6","doi-asserted-by":"publisher","DOI":"10.1109\/SPW.2018.00016","article-title":"\u201cBlack-box generation of adversarial text sequences to evade deep learning classifiers,\u201d","author":"Gao","year":"2018","journal-title":"2018 IEEE security and privacy workshops"},{"key":"B7","doi-asserted-by":"publisher","first-page":"383","DOI":"10.1080\/01621459.1974.10482962","article-title":"The influence curve and its role in robust estimation","volume":"69","author":"Hampel","year":"1974","journal-title":"J. Am. Stat. Assoc"},{"key":"B8","doi-asserted-by":"crossref","first-page":"168","DOI":"10.1145\/1014052.1014073","article-title":"\u2018Mining and summarizing customer reviews,\u201d","volume-title":"Proceedings of the tenth ACM SIGKDD international conference on knowledge discovery and data mining","author":"Hu","year":"2004"},{"key":"B9","article-title":"Bidirectional LSTM-CRF models for sequence tagging","author":"Huang","year":"2015","journal-title":"arXiv preprint arXiv:1508.01991"},{"key":"B10","doi-asserted-by":"publisher","first-page":"1248","DOI":"10.1007\/978-3-642-04898-2_594","article-title":"\u201cRobust statistics,\u201d","author":"Huber","year":"2011","journal-title":"International Encyclopedia of Statistical Science"},{"key":"B11","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2020.acl-main.197","article-title":"\u201cSMART: robust and efficient fine-tuning for pre-trained natural language models through principled regularized optimization,\u201d","author":"Jiang","year":"2020","journal-title":"Proceedings of the 58th annual meeting of the association for computational linguistics"},{"key":"B12","article-title":"Is BERT really robust? A strong baseline for natural language attack on text classification and entailment","author":"Jin","year":"2019","journal-title":"arXiv preprint arXiv:1907.11932"},{"key":"B13","doi-asserted-by":"publisher","first-page":"105227","DOI":"10.1016\/j.frl.2024.105227","article-title":"Sentiment trading with large language models","volume":"62","author":"Kirtac","year":"2024","journal-title":"Finan. Res. Lett."},{"key":"B14","doi-asserted-by":"publisher","first-page":"282","DOI":"10.1016\/j.frl.2024.105227","article-title":"\u201cConditional random fields: Probabilistic models for segmenting and labeling sequence data,\u201d","author":"Lafferty","year":"2001","journal-title":"Proceedings of the 18th international conference on machine learning"},{"key":"B15","article-title":"DP-LSTM: differential privacy-inspired LSTM for stock prediction using financial news","author":"Li","year":"2019","journal-title":"arXiv preprint arXiv:1912.10806"},{"key":"B16","doi-asserted-by":"publisher","first-page":"35","DOI":"10.1111\/j.1540-6261.2010.01625.x","article-title":"When is a liability not a liability? Textual analysis, dictionaries, and 10-ks","volume":"66","author":"Loughran","year":"2011","journal-title":"J. Finan."},{"key":"B17","doi-asserted-by":"publisher","DOI":"10.1002\/asi.23062","article-title":"Good debt or bad debt: detecting semantic orientations in economic texts","author":"Malo","year":"2013","journal-title":"arXiv preprint arXiv:1307.5336"},{"key":"B18","article-title":"Efficient estimation of word representations in vector space","author":"Mikolov","year":"2013","journal-title":"arXiv preprint arXiv:1301.3781"},{"key":"B19","doi-asserted-by":"publisher","first-page":"1979","DOI":"10.1109\/TPAMI.2018.2858821","article-title":"Virtual adversarial training: a regularization method for supervised and semi-supervised learning","volume":"41","author":"Miyato","year":"2019","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell"},{"key":"B20","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2020.findings-emnlp.341","article-title":"\u201cReevaluating adversarial examples in natural language,\u201d","author":"Morris","year":"2020","journal-title":"Findings of the association for computational linguistics: EMNLP 2020"},{"key":"B21","article-title":"\u201cA DIRT-T approach to unsupervised domain adaptation,\u201d","author":"Shu","year":"2018","journal-title":"International conference on learning representations"},{"key":"B22","doi-asserted-by":"publisher","first-page":"1314","DOI":"10.1002\/asi.24634","article-title":"SEntFiN 1.0: entity-aware sentiment analysis for financial news","volume":"73","author":"Sinha","year":"2022","journal-title":"J. Assoc. Inf. Sci. Technol."},{"key":"B23","doi-asserted-by":"publisher","first-page":"3","DOI":"10.1186\/s40537-017-0111-6","article-title":"Big data: deep learning for financial sentiment analysis","volume":"5","author":"Sohangir","year":"2018","journal-title":"J. Big Data"},{"key":"B24","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.emnlp-main.956","article-title":"\u201cFinEntity: entity-level sentiment classification for financial texts,\u201d","author":"Tang","year":"2023","journal-title":"Proceedings of the 2023 conference on empirical methods in natural language processing"},{"key":"B25","doi-asserted-by":"publisher","first-page":"1437","DOI":"10.1111\/j.1540-6261.2008.01362.x","article-title":"More than words: quantifying language to measure firms' fundamentals","volume":"63","author":"Tetlock","year":"2008","journal-title":"J. Finance"},{"key":"B26","doi-asserted-by":"publisher","DOI":"10.1109\/ACOIT62457.2024.10941671","article-title":"\u201cRobust image classification: defensive strategies against FGSM and PGD adversarial attacks,\u201d","author":"Waghela","year":"2024","journal-title":"2024 Asian conference on intelligent technologies"},{"key":"B27","doi-asserted-by":"publisher","first-page":"2526","DOI":"10.18653\/v1\/2025.acl-long.127","article-title":"\u201cSmarter, better, faster, longer: a modern bidirectional encoder for fast, memory efficient, and long context finetuning and inference,\u201d","author":"Warner","year":"2025","journal-title":"Proceedings of the 63rd annual meeting of the association for computational linguistics"},{"key":"B28","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v37i11.26608","article-title":"\u201cAdversarial self-attention for language understanding,\u201d","author":"Wu","year":"2023","journal-title":"Proceedings of the AAAI conference on artificial intelligence"},{"key":"B29","article-title":"FinGPT: open-source financial large language models","author":"Yang","year":"2023","journal-title":"arXiv preprint arXiv:2306.06031"},{"key":"B30","first-page":"7472","article-title":"\u201cTheoretically principled trade-off between robustness and accuracy,\u201d","author":"Zhang","year":"2019","journal-title":"Proceedings of the 36th international conference on machine learning"},{"key":"B31","article-title":"\u201cFreeLB: enhanced adversarial training for natural language understanding,\u201d","author":"Zhu","year":"2020","journal-title":"International conference on learning representations"}],"container-title":["Frontiers in Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2026.1888120\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,11]],"date-time":"2026-09-11T04:31:26Z","timestamp":1789101086000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2026.1888120\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,11]]},"references-count":31,"alternative-id":["10.3389\/frai.2026.1888120"],"URL":"https:\/\/doi.org\/10.3389\/frai.2026.1888120","relation":{},"ISSN":["2624-8212"],"issn-type":[{"value":"2624-8212","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,11]]},"article-number":"1888120"}}